Source location on multilayer networks

TL;DR

This paper presents a maximum likelihood method for source localization in multilayer networks, revealing counterintuitive effects where more observations can worsen results and additional layers can enhance performance.

Contribution

It extends source localization techniques to multilayer networks and uncovers novel phenomena related to observation and network complexity.

Findings

More observations can degrade localization accuracy.

Adding layers can improve source detection.

Method tested on synthetic multilayer networks.

Abstract

Nowadays it is not uncommon to have to deal with dissemination on multi-layered networks and often finding the source of said propagation can be a crucial task. In this paper we tackle this exact problem with a maximum likelihood approach that we extend to be operational on multi-layered graphs. We test our method for source location estimation on synthetic networks and outline its potential strengths and limitations. We also observe some non-trivial and perhaps surprising phenomena where the more of the system one observes the worse the results become whereas increased problem complexity in the form of more layers can actually improve our performance.